OP here. Like many of you, I saw the OpenAI Prism launch and the rumored $100M acquisition behind it. I wondered if a specialized Agentic workflow could achieve similar results without the massive model overhead.

My team and I (grad students) spent the last 24 hours hacking this together using Google's Antigravity + Gemini

The goal: Build a structured LaTeX editor that specifically handles CJK (Chinese/Japanese) logic and topology graphs better than generic models.

It's a v0.1 "weekend build" running on my GCP credits. Would love feedback on the latency vs Prism!